Open thealaskalife opened 7 years ago
I'm sure it must be a temporary thing. I only briefly tried other datasets but it should work with everything that is volumetric and has a binary label.
Fausto Milletarì Sent from my iPhone
On 15 Mar 2017, at 23:35, pcj521521 notifications@github.com wrote:
Hi, I tried to download the dataset of PROMISE12 but seems the register web is closed. Could you advise where I can download the dataset? Or any other dataset can be used to try VNet? Thanks.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.
I'm using VNet with the LUNA16 data set. I've started by just segmenting lungs which don't have a class imbalance (the CT volumes are 41% lung on average) so I swapped out Dice Coefficient for NLLoss. Adding BN and dropout I'm getting a 1.92% error rate after 10 epochs. I have code in my torchbiomed package for isotropically down scaling and padding the lung CTs and lung masks (my variant of vnet is in pytorch, but the conversion functions are framework agnostic).
Hi @pcj521521 , Currently, I can register and download the PROMISE12 data successfully. This is the register link that works for me. https://promise12.grand-challenge.org/register/
Zhuotun
The sign up link is now here: https://promise12.grand-challenge.org/participants/registration/create/
Hi, I tried to download the dataset of PROMISE12 but seems the register web is closed. Could you advise where I can download the dataset? Or any other dataset can be used to try VNet? Thanks.